Redis & Edoed
Edoed
Hey Redis, I’ve been sketching out a lightweight key‑value store that blends your data structures with a bit of optimistic concurrency control. It’s still a prototype, but I think it could help people avoid those messy lock conflicts while keeping things fast. What’s your take on adding a small, versioned persistence layer to something like this?
Redis
Sounds like a solid idea, but a versioned persistence layer will add a bit of overhead you’ll need to measure. Keep it simple – maybe just write a compact log of changes and replay it on restart. Just make sure you don’t lose the point of staying lightweight. And if you can, test the commit path under a load that simulates real contention; that’s where the optimistic part really pays off. Good luck, and remember: consistency first, convenience second.
Edoed
Thanks, I’ll keep the log tiny—just append‑only, maybe compress on the fly. I’ll stress‑test with a bursty client to see how the optimistic checks hold up under real contention. Will aim for consistency first, then trim any bloat. Catch you later with a proof‑of‑concept.
Redis
Sounds good. Just remember to keep the log size in check – you can’t afford a slow disk flush in the middle of a burst. I’ll be here when you’re ready to run it through a real‑world stress test. Good luck, and don’t let the log grow bigger than the data you’re protecting. Catch you later.
Edoed
Got it, I’ll buffer writes and sync only at safe points, maybe after a set number of entries or when the size exceeds a threshold. Will keep the log a fraction of the data size. See you when the test starts.
Redis
Buffering is the right move. Just make sure the flush interval is long enough to amortise the cost but short enough that a crash doesn’t lose too much work. Keep the threshold a bit lower than your cache size so you stay on the safe side. Good luck, and let me know when the tests are running.